Good points if you exclude validity of task-accomplishment, increasing fraud potential in teaching / research & associated time costs, dependency on non-academic AI producers or exploding energy costs for universities, & increasing over-use of natural resources from "AI's role in science".

This is exactly it: AI as a practical support. My rule for using AI in my own work is: don't use it for anything that, if time and money were no object, you wouldn't be able to do yourself. That and: don't use it for writing. I like writing. It's one of the reasons I decided to do this for a living.

LMU Professors from various disciplines share their perspectives on the role of AI in science www.lmu.de/en/newsroom/...

Depolarization is not "a scalable solution for reducing societal-level conflict.... achieving lasting depolarization will likely require....moving beyond individual-level treatments to address the elite behaviors and structural incentives that fuel partisan conflict" www.pnas.org/doi/10.1073/...

This review suggests that the average effectiveness of choice architecture interventions on behaviour is smaller than often reported and that there is substantial heterogeneity in their effects @szaszibarnabas.bsky.social www.nature.com/articles/s44...

Tidy simulation: Designing robust, reproducible, and scalable Monte Carlo simulations #StatsSky arxiv.org/abs/2509.11741 It is not formally linked to the {tidyverse}, bit affinity is obvious. The paper does a solid job in describing the simulation workflow, could be useful for intros to simulation

I regret to inform you I found *another* paper by (some of) these authors with similar errors... 🧵

The next social science revolution will be finding out that a lot of important things are non-linear.

We're looking forward to putting together a *fantastic* Electoral and Voting Behaviour section for the @epssnet.bsky.social 2026 meeting in Belfast! ‼️ Make sure to submit your papers by 7 November 2025!

Waiting for my preprint to be accepted, so in the meantime a teaser: here's what happens when you try to estimate a between-scale correlation based on LLM-generated datasets of participants, while varying 4 different analytic decisions (blue is the true correlation from human data):

Just out at JOP (www.journals.uchicago.edu/doi/epdf/10....) : Why do so few working-class people hold office? It's not lack of interest. Noam Lupu and I ran 10 surveys in 8 countries and found no evidence of a social class gap in how often qualified people consider running.

📣 MORAL APPEALS IN POLITICAL COMMUNICATION 📣 New version of @twidmann.bsky.social and my working paper answering: * Have moral appeals increased over time? * Is the tendency to moralize ideologically patterned? * Are some topics consistently more moralized than others? osf.io/preprints/os...

🚨A new dataset version of the GLES Cross-Section 2025 (ZA10100, Version 2.0.0) is available! Corrections of design weights w_ow, q37 (voting decision) ➕ Including English versions of the dataset, adjustment weights 📥 doi.org/10.4232/5.ZA...

Converse (1964)

Wow.

The good news: we found solutions that help mitigate this: ✅ Larger, more capable models are safer (but no guarantee). ✅ Few human annotations beat many AI annotations. ✅ Testing several models and configurations on held-out data helps. ✅ Pre-registering AI choices can prevent cherry-picking.

We tested 18 LLMs on 37 social science annotation tasks (13M labels, 1.4M regressions). By trying different models and prompts, you can make 94% of null results appear statistically significant–or flip findings completely 68% of the time. Importantly this also concerns well-intentioned researchers!

Long ago, in a study of the democratic attitudes in Germany, Dalton raised the possibility of “Fragebogendemokraten” (questionnaire democrats): people who hesitate to express their sincere dislike for “democracy” in surveys, providing instead the “socially desirable” response (1)

🚨 New paper alert 🚨 Using LLMs as data annotators, you can produce any scientific result you want. We call this **LLM Hacking**. Paper: arxiv.org/pdf/2509.08825

Our article “Unsuccessful Candidates Are More Concerned About Electoral Fairness than Election Winners” is now online @thejop.bsky.social Using RDD and elite survey data from Denmark, we show that losing candidates express greater concern about electoral fairness. 🔗 doi.org/10.1086/734240

John Zaller on Measurement error in standardized surveys: "Who, after all, carries around a stored attitude that ‘‘I am a 6 on the 1-to-7 Government Services scale’’? https://www.tandfonline.com/doi/full/10.1080/08913811.2012.807648?src=most-read-all-time

This is a paper I really care about. I feel the core message is very important for social scientists in general, and political scientists in particular. "Quantitative Research in Political Science is Greatly Underpowered." (with A+ co-authors)

All of us—including myself—should publish fewer but better papers. And we should spend more of our time on scrutinizing existing work, so we know which claims can be trusted. But that's hard work with little recognition. Thanks @sophieehill.bsky.social for doing it. Impressive 🕵️‍♀️ work. Fun to read.

Blog post: github.com/sophieehill/... TL;DR: There are a LOT of errors/inconsistencies in the results reported in this paper (estimates outside CIs, sign errors, duplicates, asymmetric CIs). Even in the abstract itself! This suggests manual editing of results tables. Which is not good... 🧵

Getting ready to teach internal/external/construct validity, for the first time using the framing by Estlinger, @davebrady72.bsky.social and @eschwitz.bsky.social (www.degruyterbrill.com/document/doi...). Super happy because this is much clearer than what I told student before!

This article took a really long time to be published! We started the project in 2021, then life happened. Some reviewers also didn't like the application of OSP to demoinno lit with my wonderful co-authors Matt Ryan, Rafa Mestre, Masood Gheasi and George Bolton from @sotonpolitics.bsky.social

I agree. However, Even and perhaps especially in this case a pre-registration does have value, doesn't it? Pre-registration makes the research *process* more transparent, showing in such cases that scholars did not have clear expectations that they put to a test

Pre-registered causal inference study with observational data, showing multiverse analysis in the results section. Nice!

Wir gedenken: Heute vor 25 Jahren, am 9. September 2000, wurde Enver Şimşek an seinem Blumenstand in #Nürnberg vom #NSU ermordet. Er hinterließ seine Ehefrau und seine zwei Kinder. Er war das erste Opfer der rassistischen Mordserie des NSU. #KeinSchlussstrich #KeinVergessen

Excited about the very first EPSS conference in Belfast next year!🤩 @kenbenoit.bsky.social & I co-chair the Political Communication Section and invite submissions on contents & effects of political communication, broadly conceived, and media-politics relations⬇️Please reach out in case of questions

🔍Find out about the replicability crisis across fields & open research initiatives you can implement in your own work! Attend the “Replicability Crisis” lecture by Prof. Dr. Felix Schönbrodt (@nicebread.bsky.social) on Mon 15 Sept, 9:45-10:45. 👉Register here: www.pretix.osc.lmu.de/lmu-osc/OSSS...

I am a HUGE pre-registration booster. But IMO this is the biggest problem with pre-reg (especially pre-reg without analysis plans): It's hackable if people pre-reg anything and everything. Like I've heard people *brag* about having the foresight to pre-register a lot of hypotheses. Don't do that!

Just made this little diagram and it's so pretty! {ggdiagram} is really neat for these types of graphics #rstats (code here: gist.github.com/andrewheiss/... )

Was hervorgehoben werden muss: Der neue Oberbürgermeister aus Meißen, Markus Renner, wurde im Stadtrat breit unterstützt von CDU, SPD, Grünen und Linken Mitgliedern. Das ist demokratiepolitische Verantwortung 🤝. Der Neonazi R. Jurisch wurde verhindert, ein engagierter Demokrat ist im Amt.

Radical right accommodation really does not work. New paper out with this exceptionally talented team @katharinalawall.bsky.social @robjohns75.bsky.social @drjennings.bsky.social @sarahobolt.bsky.social @zachdickson.bsky.social @danjdevine.bsky.social & @jack-bailey.co.uk doi.org/10.31235/osf...

We are about a month away from releasing a complete refresh of the OSF user interface. The team has been working on this for a very long time, and we are very excited to be able to share it soon. A preview picture:

I'm once again a session chair for APSA and will be imposing the Nyhan / Samii rules. Encourage your chairs to do the same. I don't see any good arguments for the usual all papers all discussant comments all questions format

EPSA have announced that they will hold a conference in July 2026. 😵‍💫 We understand that there might be some confusion about EPSS and EPSA. 👉🏽 So we thought we would clarify some things. A short 🧵

Enjoying the For You feed? Give it a like ♡ to help more people discover it: bsky.app/profile/did:... The more people use it -> the more feedback we get -> the better we can make it for you.

Anke Hoeffler, Tanja Kupisch, and I are happy to announce three doctoral positions in the project "Lifting the Veil of Violence: Inequality and Representation after War” (LiVV) at ‪the Cluster of Excellence @excinequality.bsky.social‬ at @uni-konstanz.de‬. stellen.uni-konstanz.de/jobposting/4...

📣 We are hiring! 👇 This PhD position is part of a research project that examines how growing income inequality can reshape the way democracy functions by influencing both political discourse and public opinion. More info here: www.academictransfer.com/en/jobs/3542...

Ever stared at a table of regression coefficients & wondered what you're doing with your life? Very excited to share this gentle introduction to another way of making sense of statistical models (w @vincentab.bsky.social) Preprint: doi.org/10.31234/osf... Website: j-rohrer.github.io/marginal-psy...

If you are preparing your bachelor statistics course and would like to add optional material for students to better understand statistics on a conceptual level (see topics in the screenshot) my free textbook provides a state of the art overview. lakens.github.io/statistical_...

New paper in AMPPS! Read this paper to learn more about #OpenScience in Majority world! Thanks to @hcp4715.bsky.social for leading this project and for amplifying the voices of the underrepresented in psychological science!

How can we achieve good measurement of attentional control? Much of the debate around this question has focused on methodological issues, most prominently the reliability paradox. Sure, reliability is important (and we critically review all recent developments).

🌅📙 64.3 Are poor people poorly heard?👂 These authors (🧵⬇️) use citizen surveys in 🇧🇪🇨🇭🇨🇦 & 🇩🇪 to claim that politicians hold biased views of what citizens want & shows that politicians perceptions parallel the rich for the most pressing economic issues. buff.ly/5UlH75b

Reminder that all three books I've co-authored are freely available online for non-commercial use (and the fourth will be, too)

There was Nothing like o3 for (private) research purposes. ChatGPT 5 (free) does not come anywhere close, even with "Deep thinking". DeepResearch seems like the only way but limited to work in a thread with multiple follow-ups.

You can already submit your paper and panel proposal for EPSS next year. Looking forward to seeing many people in Belfast. I will be there.